Microsoft also pointed out that these agents would be vulnerable to the safety threats that humans face online, like phishing, social engineering, and misinformation attacks. Also Read: Report: Salesforce Developing AI Agents that Don’t Need Human Supervision ...